https://teamchong.github.io/turboquant-wasm/draw.html
Article
-
Runs Gemma 4 entirely in-browser via WebAssembly; 3.1GB model download
-
LLM outputs compact ~50-token commands instead of raw 5,000-token Excalidraw JSON
-
Generates diagrams from natural language prompts client-side
Discussion
-
Users report impressive speed; questions about Firefox support (not supported)
-
Technical discussion: browser inference is batch-size-1, memory bandwidth bound not FLOPs
-
Requests for open source release and CDN to avoid repeated 3GB downloads
Discuss on HN